Roohi Ghosh is passionate about making the publishing journey smoother and more rewarding for researchers worldwide. As the Ambassador for Researcher Success at Cactus Communications (CACTUS), she draws on 18 years in academia to understand the hurdles that researchers face, particularly those writing in a second language. She leads initiatives in training, content development, and thought leadership, always with the goal of giving authors the tools and confidence they need to share their work with the world. Roohi’s advocacy extends beyond CACTUS: she co-chairs the Peer Review Week Committee, ALPSP’s Open Peer Review Special Interest Group, and the EASE India Chapter, while serving on several other industry committees.

This year’s theme for Peer Review Week is about rethinking peer review in the age of AI. How far have we come?

The conversation today has evolved so much. From discussions on whether AI will or will not be used in peer review, we have reached a stage where we know it is here to stay. Of course, a lot of it is experimental: AI to streamline administrative checks, detect plagiarism or image manipulation, and even suggest reviewers.

But when we really think about it, is the adoption balanced across the board? I don’t think so. Many publishers today have policies on the use of AI, but these tend to focus on what not to do, not on the responsible use of AI.

What parts of the process do you think most need to evolve?

Where we most need evolution is in three areas:

• We need clear guidelines for editors, reviewers, and authors on how AI can be used responsibly.
• We need to equip reviewers in understanding both the capabilities and the limitations of AI, so they can critically evaluate AI-assisted tools and outputs.
• We need to shift focus from speed to the value it will bring if we use AI to support peer reviewers. If the repetitive tasks are handled by AI, reviewers will be free to dedicate more of their time on deep, qualitative assessments.

At the end of the day, I think peer review in the age of AI is less about replacing humans and more about redesigning workflows and roles to combine human expertise with machine efficiency while maintaining the scientific integrity of a manuscript.        

In your view, what elements of peer review should always involve human judgment?

No matter how advanced AI becomes, elements like contextual understanding, ethical reasoning, and nuanced critique must remain human-led. Is the research novel? Is it relevant and within the journal scope? What is the impact of the research in a larger context? Are there any subtle flaws in the research methodology? Do we truly believe that AI is ready to answer these questions. I doubt it.

Moreover, sharing such feedback in a constructive and empathetic way is also something only a human can do. While AI can detect errors, it Is up to the human to mentor, to maintain scientific integrity, and to advance research. The human is irreplaceable at least for now. And I feel even as AI advances and tools become mature, the expertise, the empathy, and the ability to draw connections and interpret—these are purely human characteristics that AI cannot replace.

If at some point AI starts to become more common in peer review, how can we help reviewers use these tools responsibly and effectively?

It will take more than access to tools to help reviewers use AI responsibly. Right now, in my view, the issue is that there is no shared understanding of the reviewer role. What are the guidelines for the use of AI? Reviewers need guidance on how AI can and cannot be used. And it is the publishers who need to help create these guidelines.

Rather than issuing restrictive policies, the guidelines should focus on nuanced, use-case-based guidance and transparency requirements. Reviewers need opportunities to engage in training and dialogue. This is about trust-building: reviewers must feel confident that they can use AI without compromising integrity, and authors must trust that their work is being assessed fairly.

What role should publishers play in setting clear guidelines?

Ultimately, publishers should act as conveners—bringing together different stakeholders in academia to co-create these standards. The goal isn’t simply compliance, but a cultural shift where AI is used deliberately and ethically to strengthen, not weaken, peer review.